JOURNAL OF PETROLOGY
VOLUME 48
NUMBER 12
PAGES 2341^2357
2007
doi:10.1093/petrology/egm062
Classification of Altered Volcanic Island Arc
Rocks using Immobile Trace Elements:
Development of the Th^Co Discrimination
Diagram
A. R. HASTIE1*, A. C. KERR1, J. A. PEARCE1 AND S. F. MITCHELL2
1
SCHOOL OF EARTH, OCEAN AND PLANETARY SCIENCES, CARDIFF UNIVERSITY, MAIN BUILDING, PARK PLACE,
CARDIFF CF10 3YE, UK
2
DEPARTMENT OF GEOGRAPHY AND GEOLOGY, UNIVERSITY OF THE WEST INDIES, MONA, KINGSTON 7, JAMAICA
RECEIVED OCTOBER 3, 2006; ACCEPTED SEPTEMBER 24, 2007
ADVANCE ACCESS PUBLICATION OCTOBER 25, 2007
The recognized method of classifying most volcanic rocks
is the total alkali^silica (TAS) diagram (Le Bas et al., 1986,
1992). The total alkali (Na2O þ K2O) axis mainly distinguishes alkalic from sub-alkalic rock types, whereas the
silica axis mainly distinguishes primitive from evolved
rock types. This diagram does, however, have limited
application to volcanic arc lavas, the vast majority of
which simply classify as sub-alkaline. Volcanic arc rocks
can be further classified into tholeiitic, calc-alkaline and
shoshonitic series. At one time, this was done on the basis
of iron enrichment using diagrams such as FeOtot/MgO vs
SiO2 (Miyashiro, 1974; Arculus, 2003) or (Na2O þ K2O)^
MgO^FeOtot (the AFM plot of Kuno, 1968). Typically,
however, this subdivision is usually made using a K2O^
SiO2 diagram (Peccerillo & Taylor, 1976; Rickwood, 1989).
This diagram has the advantage of assigning both a volcanic series (tholeiitic, calc-alkaline, high-K calc-alkaline,
shoshonitic) based mainly on K2O, and hence the degree
of large ion lithophile element (LILE) enrichment, and a
rock type (basalt, basaltic andesite, andesite, dacite, rhyolite) based primarily on silica content and hence degree of
differentiation.
It has long been recognized that the TAS diagram is not
robust in classifying altered volcanic rocks, and proxies
using more immobile elements have been developed for
that purpose (e.g. Winchester & Floyd, 1977; Pearce, 1996).
However, there is no equivalent proxy for the K2O^SiO2
diagram, although there have been some efforts in that
direction (e.g. Pearce, 1982). The need for a more robust
equivalent of that diagram became apparent to us when
studying Cretaceous volcanic arc samples from Jamaica,
which have undergone hydrothermal alteration followed
*Corresponding author. Telephone: þ44 (0)29 208 75874.
E-mail: [email protected]
ß The Author 2007. Published by Oxford University Press. All
rights reserved. For Permissions, please e-mail: journals.permissions@
oxfordjournals.org
Many diagrams conventionally used to classify igneous rocks utilize
mobile elements, which commonly renders them unreliable for classifying rocks from the geological record.The K2O^SiO2 diagram, used to
subdivide volcanic arc rocks into rock type (basalts, basaltic andesites,
andesites, dacites and rhyolites) and volcanic series (tholeiitic,
calc-alkaline, high-K calc-alkaline and shoshonitic), is particularly
susceptible to the effects of alteration. However, by usingTh as a proxy
for K2O and Co as a proxy for SiO2 it is possible to construct
a topologically similar diagram that performs the same task but is
more robust for weathered and metamorphosed rocks. This study
uses 41000 carefully filtered Tertiary^Recent island arc samples
to construct aTh^Co classification diagram. A ‘testing set’comprising
data not used in constructing the diagram indicates a classification
success rate of c. 80%.When applied to some hydrothermally altered,
then tropically weathered Cretaceous volcanic arc lavas from
Jamaica, the diagram demonstrates the presence ofa tholeiitic volcanic
arc series dominated by intermediate^acid lavas overlain by a
calc-alkaline series dominated by basic lavas.
KEY WORDS: island arc lavas; element mobility; discrimination plots;
Jamaica; Caribbean
I N T RO D U C T I O N
JOURNAL OF PETROLOGY
VOLUME 48
NUMBER 12
DECEMBER 2007
Fig. 1. Simplified maps of (a) the Caribbean region, (b) Jamaica and (c) the Benbow Inlier, locating the Cretaceous arc volcanics of the Devils
Racecourse Formation. Samples collected from localities 1^5 in (c) were analysed for a full suite of major and trace elements, providing the data
reported in Table 1.
by intense tropical weathering. The aim of this study is to
devise a diagram to classify altered volcanic arc lavas on
the same basis as the K2O^SiO2 diagram and to apply it
to the Jamaican samples.
ELEMENT MOBI LITY I N
M E TA M O R P H O S E D A N D
T RO P I C A L LY W E AT H E R E D
A RC RO C K S : T H E J A M A I C A
C A S E ST U DY
The stimulus for this study was a suite of lower Cretaceous
island arc lavas from the Benbow Inlier in eastern Jamaica
(Fig. 1). The rocks are part of the Devils Racecourse
Formation, which is the oldest lava succession in Jamaica
and is composed of 1000 m of mafic and felsic lavas, volcaniclastic rocks and four interbedded limestone members
(Burke et al., 1969).
The succession was split into three sections by Burke
et al. (1969). The lower section is 400 m thick; most of the
lavas have been silicified. The top 100 m of the lower unit
consists of volcanic conglomerates (Burke et al., 1969;
Robinson et al., 1972). The middle section consists mainly
of volcaniclastic rocks, whereas the upper section is
mostly composed of pillow lavas and rare volcaniclastic
rocks (Burke et al., 1969; Roobol, 1972).
Twenty-four samples representing the succession have
been sampled and analysed for major elements and
30 trace elements by inductively coupled plasma optical
emission spectrometry (ICP-OES) and ICP mass
spectrometry (ICP-MS) using methods described by
McDonald & Viljoen (2006). These data are presented in
Table 1 together with international standard values and
analytical errors. The lavas are variably porphyritic with
plagioclase and clinopyroxene as the dominant phenocrysts, together with oxides in some samples and potassium feldspar in the more silicic rocks. The extent of
alteration is also variable, with sericite, chlorite, epidote,
calcite, clay minerals and iron oxyhydroxides as the main
secondary minerals. This reflects a hydrothermal alteration event post-dated by tropical weathering. The tropical
weathering is a particular concern for element mobility,
given the high water^rock ratios, high surface temperatures and high concentrations of organic acids
(Summerfield, 1997).
2342
HASTIE et al.
A Th-Co DISCRIMINATION DIAGRAM
Table 1: Major and trace element compositions of the Devils Racecourse lavas, Jamaica
Sample:
AHBI01
AHBI03
AHBI05
AHBI06
AHBI07
AHBI09
AHBI10
AHBI11
AHBI12
Section:
Lower
Lower
Lower
Lower
Lower
Upper
Upper
Upper
Upper
Rock type:
BA/A
BA/A
D/R
D/R
D/R
BA
BA
B
BA
Rock series:
IAT
IAT
IAT
IAT
IAT
CA
CA
CA
CA
Locality:
4
4
4
4
4
5
5
5
5
Major elements (wt %)
SiO2
5907
5898
8108
8263
7698
5002
5385
4887
TiO2
086
086
039
030
064
122
117
110
4569
115
Al2O3
1507
1504
1099
1007
1262
1702
1623
1696
1547
Fe2O3
836
839
191
092
091
1137
711
1114
1053
MnO
014
015
003
001
003
015
011
018
018
MgO
255
260
084
015
255
296
095
360
366
CaO
213
240
035
058
039
380
688
1066
946
Na2O
737
710
368
359
439
254
537
279
200
K2O
045
040
016
008
023
061
009
054
031
P2O5
013
012
006
005
014
062
029
016
025
LOI
224
231
105
069
201
786
699
243
998
Total
9838
9834
10054
9907
10090
9818
9903
9843
9867
Anhydrous SiO2
6042
6038
8194
8321
7856
5429
5789
5008
5076
Trace elements (ppm)
V
151
149
15
10
1
275
251
399
Cr
12
16
21
69
17
4
7
30
Co
187
189
234
199
Ni
9
Ga
155
Rb
Sr
375
117
38
42
5
157
97
319
138
038
37
20
111
70
020
2
8
126
189
150
43
49
58
238
435
58
115
028
387
312
5
181
497
545
362
32
243
10
155
090
467
Y
345
318
257
222
348
386
239
211
238
Zr
829
821
1132
962
1233
916
862
671
777
Nb
Ba
186
348
185
219
116
414
093
196
158
183
233
469
217
80
172
301
177
657
La
625
510
680
480
575
2300
1100
870
910
Ce
1492
1291
1692
1179
1653
4831
2794
2020
2378
Pr
206
198
247
176
285
533
412
282
365
Nd
1041
986
1138
866
1440
2390
1812
1243
1659
Sm
356
341
300
249
427
540
409
324
394
Eu
115
113
078
060
105
170
117
112
135
Gd
393
385
344
282
483
552
402
305
399
Tb
073
072
056
048
081
082
060
052
059
Dy
502
501
384
329
570
546
395
341
387
Ho
102
102
075
064
111
106
075
064
072
Er
304
306
241
211
365
337
239
185
227
Tm
046
045
038
033
060
051
037
028
035
Yb
315
323
255
214
394
326
239
192
217
Lu
049
051
046
040
065
054
040
030
042
Hf
216
222
312
265
338
263
237
152
214
Ta
007
008
011
009
011
014
012
006
011
Th
035
034
083
069
065
171
162
117
131
U
028
028
050
040
057
048
096
062
059
(continued)
2343
JOURNAL OF PETROLOGY
VOLUME 48
NUMBER 12
DECEMBER 2007
Table 1: Continued
Sample:
AHBI13
AHBI14
AHBI15
AHBI16
AHBI17
AHBI18
AHBI19
AHBI20
AHBI21
Section:
Upper
Upper
Upper
Lower
Lower
Lower
Lower
Lower
Lower
Rock type:
B
BA
BA
BA
D
BA/A
D
D/R
D/R
Rock series:
CA
CA
CA
IAT
IAT
IAT
IAT
IAT
IAT
Locality:
5
5
5
2
2
2
2
2
2
Major elements (wt %)
SiO2
5249
4612
4201
5398
5944
6480
6582
7573
7701
TiO2
120
117
111
117
097
127
077
039
040
Al2O3
1611
1469
1753
1637
1467
1157
1567
1133
951
Fe2O3
1112
1383
916
1203
939
985
669
341
456
MnO
013
024
015
016
020
025
013
003
008
MgO
265
249
307
335
039
229
228
069
110
CaO
385
681
986
495
468
227
033
153
050
Na2O
575
428
412
465
512
453
524
193
332
K2O
025
153
048
018
019
006
005
213
005
P2O5
024
024
024
015
017
025
013
005
008
LOI
581
695
1120
297
354
248
262
196
233
Total
9959
9835
9892
9996
9875
9963
9972
9918
9893
Anhydrous SiO2
5572
4957
4731
5563
6162
6645
6760
7725
7885
Trace elements (ppm)
V
387
374
325
370
341
320
84
33
15
Cr
12
125
58
23
19
25
29
34
17
Co
304
217
311
102
229
Ni
5
43
34
18
39
32
32
Ga
161
186
116
108
169
107
128
Rb
Sr
227
525
250
30
179
1448
490
155
205
—
104
689
122
328
389
052
112
96
051
62
35
1029
98
47
012
32
Y
349
229
271
251
214
337
333
424
397
Zr
894
843
697
499
289
592
741
1882
1033
Nb
Ba
214
306
200
433
178
143
125
096
191
71
156
99
156
58
457
665
187
29
La
1450
1100
1325
308
370
424
417
939
654
Ce
3543
2590
2601
786
824
1053
1033
2212
1553
Pr
530
352
401
146
110
192
186
358
262
Nd
2331
1524
1698
756
578
958
906
1603
1264
Sm
550
376
417
260
200
316
299
473
392
Eu
174
130
139
100
080
113
092
101
100
Gd
499
351
420
324
271
395
366
509
455
Tb
083
062
067
059
047
072
067
094
084
Dy
532
390
417
389
316
473
452
635
558
Ho
101
075
084
078
064
095
096
131
118
Er
295
225
241
229
187
279
291
403
359
Tm
043
034
036
037
030
045
046
067
059
Yb
281
224
238
238
190
281
303
464
401
Lu
045
035
037
035
032
043
047
073
061
Hf
207
203
184
154
091
168
226
542
290
Ta
007
007
010
007
005
009
011
033
013
Th
143
141
144
017
009
022
032
150
059
U
110
051
075
013
010
015
024
070
050
(continued)
2344
HASTIE et al.
A Th-Co DISCRIMINATION DIAGRAM
Table 1: Continued
Sample:
AHBI22
AHBI23
AHBI26
AHBI27
AHBI28
Section:
Lower
Lower
Lower
Lower
Lower
AHBI30
Lower
Rock type:
D/R
BA/A
D/R
D/R
D/R
D
Rock series:
IAT
IAT
IAT
IAT
IAT
IAT
Locality:
2
3
3
3
3
1
Major elements (wt %)
SiO2
8142
5699
7767
7322
7634
TiO2
031
102
043
053
044
6609
050
Al2O3
816
1459
1057
1279
1059
1533
Fe2O3
149
1022
288
476
414
540
MnO
002
012
005
004
007
009
MgO
036
192
043
090
065
244
CaO
085
745
114
020
024
311
Na2O
254
447
399
518
473
290
K2O
029
003
006
006
004
072
P2O5
005
012
009
009
009
010
LOI
098
303
168
198
155
273
Total
9647
9995
9899
9977
9889
9940
Anhydrous SiO2
8223
5877
7899
7470
7754
6795
Trace elements (ppm)
V
26
307
8
6
5
Cr
45
44
18
39
39
Co
Ni
24
67
Ga
61
Rb
104
Sr
50
189
24
40
34
104
9
104
31
23
66
18
21
242
120
176
131
149
043
78
025
064
56
28
048
24
294
136
Y
228
221
404
495
346
185
Zr
972
333
1087
1457
1179
742
Nb
Ba
134
211
097
8
209
271
264
43
208
19
171
700
La
435
280
673
731
584
455
Ce
925
660
1602
1748
1514
1007
Pr
149
117
282
311
263
152
Nd
670
610
1378
1447
1277
659
Sm
209
214
417
462
398
186
Eu
055
091
111
129
086
067
Gd
245
275
488
565
456
208
Tb
048
049
090
109
082
039
Dy
324
329
596
739
547
265
Ho
070
068
123
153
111
055
Er
217
199
367
464
337
173
Tm
036
031
057
077
055
030
Yb
251
204
369
510
363
202
Lu
041
032
053
074
059
033
Hf
268
107
328
414
343
204
Ta
010
006
015
017
015
013
Th
074
011
064
079
066
076
U
037
012
051
078
048
046
(continued)
2345
JOURNAL OF PETROLOGY
VOLUME 48
NUMBER 12
DECEMBER 2007
Table 1: Continued
JB-1a certified values
Av. JB-1a value for this
RSD
Detection limits
analysis
Major elements (wt %)
SiO2
5216
5280
086
00119
TiO2
130
128
256
00002
Al2O3
1451
1471
145
00055
Fe2O3
910
895
165
00044
MnO
015
015
361
00194
MgO
775
794
124
00004
CaO
923
952
207
00029
Na2O
274
258
688
00029
K2O
142
137
696
00169
P2O5
026
026
284
00044
Trace elements (ppm)
V
206
196
225
007
Cr
415
423
426
021
Co
395
381
331
003
Ni
140
139
723
034
Ga
180
179
189
0022
Rb
14
14
9124
0031
Sr
443
466
932
029
Y
240
240
158
002
Zr
1460
1375
486
005
Nb
27
28
546
009
Ba
497
502
218
041
La
381
376
365
0011
Ce
661
664
188
0006
Pr
73
72
214
0003
Nd
255
258
152
0006
Sm
507
512
353
0005
Eu
147
149
144
0002
Gd
454
460
331
0028
Tb
069
068
330
0009
Dy
419
406
167
0003
Ho
072
074
265
0001
Er
218
213
259
0003
Tm
031
031
265
0001
Yb
21
207
238
0003
Lu
032
031
644
0004
Hf
348
336
512
0002
Ta
16
16
241
0001
Th
88
88
328
0002
U
16
16
915
0004
Major elements determined by ICP-OES; trace elements by ICP-MS. Rock types: B, basalt; BA, basaltic andesite;
A, andesite; D, dacite; R, rhyolite. Rock series: IAT, island arc tholeiite; CA, calc-alkaline. RSD, relative standard
deviation. The locality numbers refer to Fig. 1. The JB-1a international standard data are based on 10 analyses carried out
within the Jamaica analytical run. Detection limits are based on blank analyses. The analytical procedure has been
described by McDonald & Viljoen (2006). JB-1a certified values are taken from Govindaraju (1994). LOI, loss on ignition.
2346
HASTIE et al.
A Th-Co DISCRIMINATION DIAGRAM
The behaviour of an element during weathering and
hydrothermal alteration is commonly related to its
charge/radius ratio (ionic potential) (e.g. Pearce, 1996).
Thus, elements that form ions of low ionic potential
(5003 pm1) tend to be preferentially removed in solution
as hydrated cations, whereas elements forming ions with a
high ionic potential (4010 pm1) tend to be preferentially
removed as hydrated oxyions. Ions of intermediate ionic
potential (003^010 pm1) tend to remain in the solid product of weathering and so are relatively immobile; this is
typically true even at greenschist-grade metamorphism.
Thus, the elements Zr, Hf, Nb, Ta, Y, Ti, Cr, the rare
earth elements (REE) apart from Eu and possibly La, Th,
Ga and Sc are among the most immobile. However, a
change in fluid composition from H2O-rich to a CO2-rich
fluid or SiO2-rich partial melt, and/or an increase in temperature and/or extremely high fluid throughput may
mobilize otherwise immobile elements (e.g. Hynes, 1980;
McCulloch & Gamble, 1991; Pearce, 1996; Hill et al., 2000).
The mobility of elements in each case must therefore be
tested. An effective method is that of Cann (1970), in
which an immobile element is plotted on the horizontal
axis of bivariate variation diagrams and elements to be
evaluated are plotted on the vertical axes. If the two elements are moderately highly incompatible and immobile,
and the samples are cogenetic, the data should give trends
with slopes close to unity.
For this study, Nb is used as the immobile element
(Fig. 2) because it is one of the most immobile (e.g. Cann,
1970; Hill et al., 2000; Kurtz et al., 2000) and has a similar
partition coefficient to Th, one of the key elements considered here. Plotting Zr against Nb in Fig. 2a gives a linear
trend with a slope of unity for both lower and upper Devils
Racecourse lavas, indicating that both elements are
immobile and that the variations may be explained by
intra-formation differentiation. Figure 2b similarly exhibits
significant (95% or better) within-formation correlations,
providing evidence for Yb immobility. It also allows the
lower and upper units to be distinguished based on their
Nb/Yb ratios.
In Fig. 2c, Th similarly exhibits immobile behaviour
with two distinct trends, one with high Th/Nb ratios and
one with lower ratios. In contrast, in Fig. 2d, U (which
is known to be mobile during oxidative alteration) predictably gives a much greater scatter. In Fig. 2e and f, the light
REE (LREE), La, and middle REE (MREE), Sm, also
give inter-formation correlations consistent with immobility, Sm being more immobile than La.
Significantly for this work, the elements of low ionic
potential known to be mobile in most settings (exemplified
here by Ba and K), exhibit a large scatter with no evidence
of the expected, pre-alteration slope of unity within the
lower and upper Devils Racecourse lavas. Clearly, therefore, K has been variably added and/or subtracted during
hydrothermal alteration and weathering and is not useful
for classifying these rocks; moreover, of the LILE used to
identify rock series, Th is likely to be the most robust,
although LREE may also be usable in this case.
F I N G E R P R I N T I N G VO L C A N I C
A RC RO C K T Y P E U S I N G
PU BLISH ED I M MOBI LE
E L E M E N T D I AG R A M S
Many of the earliest immobile trace element classification
diagrams, such as those of Pearce & Cann (1973), Wood
et al. (1979) and Shervais (1982), focused on fingerprinting
the tectonic setting of volcanic rocks. Jackson (1987) used
these diagrams to assign an island arc origin to the Devils
Racecourse Formation. The normal mid-ocean ridge
basalt (N-MORB)-normalized trace element patterns in
Fig. 3a clearly highlight the arc-like negative Nb anomalies
that characterize all the samples.
Only Floyd & Winchester, in a series of papers (e.g.
Floyd & Winchester, 1975, 1978; Winchester & Floyd, 1976,
1977), specifically addressed the identification of rock type.
The most commonly used approach is their Zr/TiO2^Nb/
Y diagram (Winchester & Floyd, 1977), which has subsequently been updated using a much larger dataset and
statistically drawn boundaries by Pearce (1996) (Fig. 3b).
This diagram is essentially a proxy for the TAS classification diagram, where Nb/Y is a proxy for alkalinity
(Na2O þ K2O) and Zr/TiO2 is a proxy for silica. Nb/Y
increases from sub-alkalic to alkalic compositions and
Zr/TiO2 increases from basic to acid compositions. The
Jamaican samples from the Lower Formation give a wide
spread from basalts to dacites or rhyolites, and the Upper
Formation plots entirely as basalts (Fig. 3b).
Diagrams such as Fig. 3b, however, have limitations in
classifying volcanic arc rocks, as the original Winchester
& Floyd (1977) diagram was constructed using all types of
igneous rocks except for island arc lavas. This was because
of the limited number of analyses of volcanic arc rocks in
the 1970s. Thus, the diagrams classify volcanic rocks on the
basis of their sub-alkaline and alkaline characteristics but,
unlike diagrams such as the Th/Yb^Ta/Yb diagram,
cannot subdivide rocks into more specific tholeiitic, calc-alkaline, high-K calc-alkaline and shoshonitic fields
(Fig. 3b).
The updated Zr/TiO2^Nb/Y diagram of Pearce (1996)
uses volcanic arc analyses but it raises another problem;
namely, the large overlap displayed by island arc basalts,
basaltic andesites, andesites and dacites. This overlap is
partly because the high water contents of volcanic arc
magmas depolymerize the melt creating oxidizing conditions, which causes Fe^Ti oxides to crystallize earlier than
in magmas from other tectonic settings. Because melt
water content varies between arc volcanoes, Fe^Ti oxide
2347
JOURNAL OF PETROLOGY
VOLUME 48
NUMBER 12
DECEMBER 2007
Zr ppm
Yb ppm
100
(a)
(b)
U ppm
Th ppm
10
(d)
La ppm
Sm ppm
(c)
(e)
Ba ppm
K2O wt%
(f)
(h)
(g)
Nb ppm
Nb ppm
Fig. 2. Variation diagrams for a range of elements plotted against the most immobile element, Nb. In all diagrams, variations within the lower
and upper Devils Racecourse Formation at basic^intermediate compositions are mainly due to fractional crystallization, which should give
near-diagonal (1:1) vectors on log^log plots. Acid rocks have more complex petrogeneses and have not been plotted.
2348
HASTIE et al.
A Th-Co DISCRIMINATION DIAGRAM
100
crystallization, which contributes towards the increase in
the silica content of the residual melts, begins at different
Zr concentrations and Zr/TiO2 ratios. In addition,
because Zr is more incompatible than Ti, the Zr/TiO2
ratio is influenced by the degree of partial melting and
the processes that cause mantle heterogeneity; it is therefore variable in arc magmas before fractional crystallization takes place. This is true of any ratio using elements of
different compatibilities or any element that is moderately
to highly incompatible.
To address the problem that the term ‘sub-alkaline’ does
not fully describe all volcanic arc lavas, Pearce (1982) used
an ‘X’/Yb^Ta/Yb diagram not only to fingerprint arc
lavas, but also to identify the volcanic series. The concept
is that, in the subduction environment, some elements
(including Ta and Yb) will remain in the slab and may be
described as ‘conservative’ whereas others are transferred
to the mantle wedge by fluids and/or melts and may be
described as ‘non-conservative’ (Pearce & Peate, 1995).
If X is a non-conservative element, volcanic arc data will
lie above the MORB array on this type of projection.
In addition, the degree of displacement from the MORB
array increases from tholeiitic though calc-alkaline to
shoshonitic compositions (e.g. Fig. 3c for X ¼ Th). For the
Jamaican rocks plotted on this diagram, the Lower
Formation plots in the island arc tholeiite field and the
Upper Formation plots in the calc-alkaline field (Fig. 3c).
However, the use of trace element ratios means that the
diagram is largely independent of the effects of fractional
crystallization and is able to identify rock series but not
rock type. It is thus not the proxy for the K2O^SiO2
diagram that is sought here.
(a)
Rock/N-MORB
10
1
0.1
Th Nb La Ce Pr Nd Sm Hf Zr Eu Ti Gd Tb Dy Y Ho Er Tm Yb
alkali
rhyolite
(b)
rhyolite
dacite
evolved
phonolite
trachyte
nolite
Zr/TiO2
ripho
y
trach e
sit
ande
e
site
ande ic andesit
lt
basa
intermediate
teph
foidite
alkali
basalt
basic
basalt
sub-alkaline
alkaline
ultra-alkaline
Nb/Y
(c)
shoshonite
Th/Yb
calc-alkaline
C HOIC E OF ELEM EN TS
F O R A K 2 O ^ S I O 2 P ROX Y
D I AG R A M
For a new classification diagram to be constructed,
the mobile elements K and, to a lesser extent, Si have
to be replaced with immobile elements that behave in
a similar way during subduction zone processes, but subsequently remain immobile during surface weathering. The
selection of suitable proxy elements requires knowledge of
how these elements behave during subduction zone
processes.
Lower devils Racecourse formation
Upper devils Racecourse formation
y
rra
-a
islandarc
tholeiite
B
OR
M
Ta/Yb
Replacement of K by a relatively
immobile element
Fig. 3. (a) N-MORB normalized multi-element patterns (given as
ranges) highlighting the arc-like character (large negative Nb anomalies) of the lavas of the Lower and Upper Devils Racecourse
Formation. Normalizing values in (a) are from Sun & McDonough
(1989). (b) Zr/TiO2^Nb/Y diagram of Pearce (1996) (after Winchester
& Floyd, 1977), which is widely used as an immobile element proxy for
the TAS diagram. (c) Th/Yb^Ta/Yb ratio plot of Pearce (1982), which
provides an immobile element method of identifying arc lavas and
their volcanic series. Key as in Fig. 2.
The difficulty in finding immobile elements to classify
arc volcanic rocks has always been the fact that nonconservative elements have to be mobile at some stage to
be driven off the subducting plate. It is now well known
from experiment and observation that elements are transferred from the subducting plate to the mantle wedge over
different temperature ranges. As the plate subducts, the
most fluid-mobile elements such as B and Cs begin to be
released at the shallowest depth, then the typical fluidmobile elements such as K and Ba, and finally the least
2349
JOURNAL OF PETROLOGY
VOLUME 48
fluid-mobile elements, such as Th and the LREE, begin to
be released at still greater depth (e.g. Becker et al., 2000;
Savov et al., 2005). This last group provides the only proxies
for K in terms of immobility, although their different
release profiles mean that they do not behave in precisely
the same way. This is apparent in Fig. 2, where Th proves to
be the most immobile element that also behaves as a nonconservative element in a subduction context. Thus Th has
been chosen to substitute for K in the new classification
diagram.
In detail, the mechanism for transport of Th is controversial. A number of studies have used Pb isotopes and
other parameters to distinguish two main subduction components: (1) aqueous fluid derived mainly from altered
oceanic crust; (2) silicate melts derived mainly from sediment (e.g. Miller et al., 1994; Brenan et al., 1995; Regelous
et al., 1997; Class et al., 2000; Elliott, 2003). It is clear that
K is mobile in both components but less clear how Th
behaves. One view is that it is mobile only in a sediment
melt component (e.g. Elliott, 2003), and it is true that
experiments indicate a marked increase in Th mobility
when the sediment solidus is reached (Johnson & Plank,
1999). On the other hand, these experiments indicate that
Th is still mobile, although less so, under subsolidus conditions above 6008C, raising the possibility that Th is also
transported by supercritical, subsolidus aqueous fluids.
This view has been supported by Keppler (1996), Kessel
et al. (2005) and others, who have presented convincing
experimental evidence that both a ‘normal’ and a chloriderich fluid from a subducted slab would readily transport
Th into the overlying mantle wedge. However, they too
demonstrate decreasing Th mobility with falling temperature. Thus for warm to cool subduction, it is likely that only
K is transported by crust-derived fluids at shallow depth,
whereas Th and K are transported by sediment melt at
greater depth. Between these depths, there is a depth
range marked by transport of fluids containing both K
and Th but with high K/Th. This is not a problem for the
classification provided the shallow and deep subduction
components are integrated before they contribute to arc
magma genesis.
Viewed from an empirical perspective, various studies
have demonstrated Th immobility during tropical weathering: indeed, the immobility of Th in contrast to the
mobility of U is the basis for U^Th disequilibria investigations of weathering rates (e.g. Dosseto et al., 2006).
Th mobility in metamorphosed crust probably begins in
the upper amphibolite facies, so one might infer that it
becomes significant at temperatures between about 450
and 6508C. In fingerprinting rocks, this means that it can
certainly be treated as immobile in rocks of greenschist
facies and below. It also means that a large proportion of
the subducted Th will be transmitted to the mantle wedge
NUMBER 12
DECEMBER 2007
by the time sub-arc depths are reached. Thus, for the most
part, Th can be a proxy for K, even though there are
differences in detail in some cases. This is evident in
Fig. 4a, where K and Th exhibit near-linear trends in the
fresh rocks of modern arcs (Honshu and the Aleutians are
shown), even though these may be made up of a range of
volcanoes from different locations with respect to the
Benioff Zone. However, there is a small component of variation orthogonal to the main trends. This may represent
partial decoupling of K and Th during subduction, partial
melting effects (K is less incompatible than Th) or some
mobilization of K by volcanic fluids. Figure 4a also
demonstrates that K mobility causes such relationships to
break down entirely in altered equivalents, such as the
lavas from Jamaica.
It should be noted that Th/Yb (Fig. 3c) can also be used
as a measure of LILE enrichment to discriminate between
tholeiitic, calc-alkaline and shoshonitic compositions.
There are both benefits and disadvantages in using such a
ratio; the advantages are that it reduces the effects of fractional crystallization, plagioclase accumulation and concentration by leaching or dilution by veining. However,
the aim here is to provide a proxy for the K2O^SiO2
diagram of Peccerillo & Taylor (1976). For that purpose,
it is the element that has to replace K2O, not the ratio.
Replacement of SiO2 by a relatively
immobile element
The most successful immobile element proxy for silica to
date has been Zr/TiO2 (e.g. Winchester & Floyd, 1977).
The basis of this ratio is that it mirrors silica by changing
little during crystallization of olivine, pyroxene and plagioclase (i.e. within basalts), but increases when Fe^Ti oxides
crystallize and help drive the magma composition to more
SiO2-rich compositions. Unfortunately, as noted above,
the proxy is only partially effective for volcanic arcs.
In terms of partition coefficients, Si is an element that is
slightly incompatible throughout the fractional crystallization and assimilation history of most sub-alkaline magmas.
There are no precise immobile element equivalents that we
are aware of. Therefore, instead of using an incompatible
element we have chosen to use a compatible element,
which will be gradually removed from the melt throughout
the crystallization sequence and so will reflect fractionation from basalt to rhyolite in an inverse way to silica.
Of the possible candidates, Ni and Cr are too compatible, commonly having values below detection limits in
intermediate^acid magmas. Sc is incompatible during
melting and slightly to moderately compatible during
fractional crystallization, but primary magmas have very
variable Sc concentrations because of the variable role of
garnet in their genesis (e.g. Pearce & Parkinson, 1993).
The most effective trace element, both from a theoretical
standpoint and from a series of empirical tests, is Co.
2350
HASTIE et al.
A Th-Co DISCRIMINATION DIAGRAM
100000
Aleutian arc
Honshu arc
Lower devils Racecourse
Upper devils Racecourse
K ppm
10000
1000
(a)
100
0.1
1
10
Th ppm
(b)
(c)
Fig. 4. (a) K^Th, (b) Co^SiO2 and (c) Co^Zr/TiO2 variation diagrams for evaluating the immobile element proxies (Th for K, Co and
Zr/TiO2 for SiO2). The diagram illustrates trends from some typical
fresh arc lavas from the Aleutian and Honshu arcs from Nakano (1993,
1994) and Hildreth et al. (2004) for comparison with data from the
highly altered Jamaica lavas.
Cobalt is partitioned strongly into olivine, which buffers its
concentration during partial melting, ensuring little variation between primary volcanic arc magmas. During fractional crystallization, it is highly compatible in olivine and
Fe^Ti oxides, slightly compatible with respect to pyroxenes
and amphibole, and strongly incompatible with respect to
feldspar (Pearce & Parkinson, 1993). For most mineral
assemblages, the bulk distribution coefficient is 15^2. The
Co^SiO2 variation diagram illustrated in Fig. 4b, again
using data from the Aleutian and Honshu arcs, confirms
the highly significant inverse correlation between the two
elements, providing the basis for Co to be used as a proxy
for SiO2.
For the Jamaican samples, Fig. 4b also shows a good
correlation but close examination reveals potential issues.
In particular, the Upper Devils Racecourse Formation
has a wide range of silica contents (c. 42^54 wt %, or
47^58 wt % anhydrous) for little change in Co that would
imply a range of rock types from basic to intermediate.
However, it is apparent from the Zr/TiO2^Nb/Y diagram
(Fig. 3b) that these samples form a tight cluster, implying
that they all belong to a similar rock type. Thus, most of
the silica variation apparent in Fig. 4b must be a function
of alteration. Using SiO2 contents recalculated on an anhydrous basis increases the concentrations but retains the
variance.
The Co^Zr/TiO2 diagram (Fig. 4c), comprising the two
immobile element proxies, is also informative. Here the
Honshu and Aleutian arc examples both form linear
trends, the Honshu arc displaced to lower Zr/TiO2 for
a given Co concentration. The Jamaica samples are displaced to still lower Zr/TiO2 concentrations, probably
reflecting the more depleted nature of the source of their
primary magmas. The uniform composition of the Upper
Devils Racecourse Formation is apparent and alhough a
broad negative correlation is evident in the Lower Devils
Racecourse data there is more scatter than in the
Aleutian and Honshu arc data. In particular, there are
two additional trends: one to low Zr/TiO2 in the more
basic samples, and one to low Zr/TiO2 in the most acid
samples. Given that the former samples have Zr/TiO2
lower than any normal basalts, we interpret this trend in
terms of oxide accumulation. The second trend reflects a
decrease in Zr concentration in the most silicic rocks and
we interpret this in terms of zircon fractionation. Neither
oxide nor zircon saturation in the melt has a major
impact on Co, making Co potentially a better SiO2 proxy
than Zr/TiO2. Co does have one potential cause for concern, however: it will increase markedly in response to
olivine accumulation. None the less, olivine accumulation
also reduces SiO2, so the olivine accumulation vector is
only slightly steeper than the magmatic Co^SiO2 trend
and presents less of a problem for classification.
2351
JOURNAL OF PETROLOGY
VOLUME 48
The immobility of Co is not immediately obvious. Co
has two oxidation states, Co2þ and Co3þ having Z/r of
0276 and 0476 pm^1, respectively, with the result that the
former, and dominant, species is potentially mobile and
the latter is not. However, one of the principal characteristics of Co that limits its mobility in tropical weathering is
its strong adsorption onto iron and manganese oxyhydroxides with accompanying oxidation of Co2þ to Co3þ.
Laterite profiles in ultramafic terranes have long demonstrated that Co immobility accompanies extensive silica
loss during tropical weathering (e.g. Trescases, 1973), so Co
is an ideal element with which to study the weathered arc
lavas from Jamaica. During metamorphism, Co compatibility in Fe-rich minerals such as chlorite, Fe^Ti oxides
and amphibole similarly restricts Co mobility. Thus,
although care is needed to confirm immobility, Co should
be effective in the classification of most altered and metamorphosed rocks.
T H E T h ^ C o C L A S S I F I C AT I O N
D I AG R A M
Compositions of relatively young island arc lavas were
taken from the Earth Reference Data and Models database (http://www.earthref.org). The new Th^Co classification diagram was constructed using data from the
Tertiary^Recent Aeolian, Aleutian, Andean, Banda,
Central American, Honshu, Izu^Bonin, Kamchatka,
Kermadec, Lesser Antilles, Luzon, Mariana, New
Hebrides, New Zealand, Papua New Guinea, Ryukyu,
Sunda and Tonga arcs. Samples designated in the database
as being ‘weathered’ or ‘metamorphosed’ were removed, as
were those with other indices of potential K mobility such
as abnormally high water contents or high loss on ignition
values. Adakites, boninites, xenoliths, nephelinites and
picrites were similarly removed from the dataset to ensure
that only basalts, andesites, dacites and rhyolites were used
to construct the new classification diagram. Data clearly
below cited or estimated analytical resolution limits were
also removed. This left 1095 samples with data for both
Th and Co that could be used to construct the classification diagram (references are given in Electronic
Appendix A, which is available for downloading from
http://petrology.oxfordjournals.org). This is the ‘training
set’ (Pearce, 1976).
A second subset of data from two arcs not used in setting
up the classification diagram was ‘held back’ to independently test the effectiveness of the method. This subset is
from the Bismarck, Kurile and Aegean arcs and represents
the ‘testing set’ (references are available in Electronic
Appendix B, which is available for downloading from
http://petrology.oxfordjournals.org).
It should be noted that, despite our efforts to filter the
data, the data quality from the Earth Reference Data and
NUMBER 12
DECEMBER 2007
Models Database is a slight concern for this study.
Predominantly, the filtered Co and Th data are from
three analytical methods: more recent ICP-MS, older
instrumental neutron activation analysis (INAA), and
(for high Th only) X-ray fluorescence (XRF). These methods usually yield reliable data, but we accept that full quality control was not available for all data and thus that the
dataset is not internally consistent.
The first step in the construction of the new classification
diagram was to plot all the data on the Peccerillo & Taylor
(1976) K2O^SiO2 diagram, thus allowing all lavas to be
labelled according to their degree of differentiation and K
enrichment. The distribution of data, plotted to separately
emphasize rock type and rock series, is illustrated in Fig. 5.
The arc data, classified on the basis of K2O and SiO2 in
Fig. 5, were then plotted on the Th^Co diagram (Fig. 6).
For each rock type and series, percentage contours were
used to construct field boundaries. The contours can be
drawn around either the mean or the peak position of the
distribution of the data (Le Bas et al., 1992; Pearce, 1996).
The Th^Co diagram contained some anomalous samples
and therefore the mean values gave erroneous results.
Thus, percentage contours were constructed around
the peak distribution positions (Fig. 6). Pearce (1996)
used the 90% contour to revise the Zr/TiO2^Nb/Y
diagram of Winchester & Floyd (1977), whereas Le Bas
et al. (1992) used the 75% contour to construct the fields
on the TAS diagram. We have arbitrarily chosen the 85%
contour for each magma type on the Th^Co diagram
(Fig. 6).
The success of the Th^Co diagram as a proxy for the
Peccerillo & Taylor (1976) K2O^SiO2 diagram can be
demonstrated by classifying the samples used to devise the
diagram (i.e. the ‘training set’). Table 2 gives the results.
Classifying according to magma series gives a success rate
just above 80% for tholeiitic and calc-alkaline samples.
However, the boundary between the calc-alkaline and the
high-K calc-alkaline and shoshonitic lavas has a lower
success rate of 78%. In part, this is because of the lack of
samples and the lack of a reliable distribution peak. In particular, it should also be noted that the numbers of samples
for trachyte and latite in the database were extremely
small, which may necessitate revision of these fields as
more data become available. Classifying according to
rock type (basalt, basaltic andesite^andesite, dacite^
rhyolite) gave an average success rate of 77%, the biggest
overlap being between basalt and basaltic andesite^
andesite rock types. Table 2 also shows that the ‘testing set’
of samples from the Bismarck, Kurile and Aegean island
arcs has an overall success rate of 480% for both the
magma series and rock type classification, similar to that
of the ‘training set’.
There are many reasons why the success rate is not greater
than 80%. Analytical errors in the determination of Th and
2352
HASTIE et al.
A Th-Co DISCRIMINATION DIAGRAM
100
IAT
CA
H-K
SHO
Th ppm
10
1
0.1
(a)
(a)
0.01
100
Th ppm
10
1
0.1
(b)
(b)
Fig. 5. K2O^SiO2 discrimination diagrams for the 1095 island arc
analyses (Electronic Appendix A). Compositional fields are from
Peccerillo & Taylor (1976). IAT, island arc tholeiite; CA, calc-alkaline;
H-K, high-K calc-alkaline; SHO, shoshonite. (a) Rock series classification and (b) rock type classification for comparison with Fig. 6.
Co probably account for a significant proportion of
the ‘missing’ 20%. Also, as noted above, Th is not a
perfect proxy for K, as it is usually not mobilized in a subduction zone at shallow depths. In addition, there are other complicating factors. For example, Co and SiO2 are concentrated
in different minerals and so are affected differently by crystal
accumulation. None the less, it is unlikely that any diagram
could precisely reproduce the K2O^SiO2 diagram and, as
new and better data become available, it should be possible
to improve theTh^Co diagram (Fig.7).
Applying the diagram to the Jamaican samples reveals a
change in arc chemistry from tholeiitic to calc-alkaline
affinities. The Th^Co diagram indicates that the lowermost lavas have tholeiitic affinities and range in composition from basaltic andesites to dacites^rhyolites. The
uppermost lavas have calc-alkaline affinities and range in
composition from basalts to basaltic andesites (Fig. 8).
0.01
70
60
50
40
30
Co ppm
20
10
0
Fig. 6. Th^Co discrimination diagrams showing the 1095 arc lavas
classified using Fig. 5. (a) Samples plotted as rock series; (b) samples
plotted as rock type. Key as in Fig. 5. The 85% contour lines are also
illustrated for each rock series and rock type.
These results are very similar to those obtained based on
Fig. 3b and c, but now they may be achieved slightly
more precisely and by a single proxy diagram rather than
two diagrams.
CONC LUSIONS
2353
(1) Because of the near-ubiquitous alteration and/or
weathering of ancient volcanic rocks, it is useful to
have immobile element proxy diagrams to replace
conventional diagrams for rock classification. At present, at least one diagram (Zr/TiO2^Nb/Y) can be
used to replace the total alkali^silica (TAS) diagram.
However, there is no immobile element equivalent for
JOURNAL OF PETROLOGY
VOLUME 48
NUMBER 12
DECEMBER 2007
Table 2: Classification of the training set (a) and the testing set (b) samples using theTh^Co discrimination diagram
(a) Classification of the training set
Th–Co magma type
IAT
Sum
CA
% correct
HK-SHO
SiO2–K2O magma type
IAT
193
45
1
239
81
CA
63
481
30
574
84
3
83
196
282
H-K/SHO
Average% correct
70
78
Average% correct for IAT and CA
82
Th–Co magma type
Basic
Sum
Intermediate
% correct
Acid
SiO2–K2O magma type
Basalt
375
56
3
434
86
Basaltic andesite–andesite
138
366
29
533
69
5
26
97
128
76
Dacite–rhyolite
Average% correct
77
(b) Classification of the testing set
Th–Co magma type
IAT
Sum
CA
% correct
HK-SHO
SiO2–K2O magma type
IAT
38
10
0
48
79
CA
5
28
2
35
80
H-K/SHO
0
0
11
11
100
Average% correct
86
Th–Co magma type
Basic
Sum
Intermediate
% correct
Acid
SiO2–K2O magma type
Basalt
26
8
0
34
76
Basaltic andesite–andesite
11
52
1
64
81
0
2
22
24
92
Dacite–rhyolite
Average% correct
83
In (a) and (b) the ‘sum’ columns denote the number of samples classified with the K2O–SiO2 diagram of Peccerillo
& Taylor (1976). Data in the other columns are the number of samples classified in each category using the Th–Co
diagram. The samples that classify correctly on the Th–Co diagram are also given as a percentage. The ‘training set’ is the
data used to devise the diagram; ‘the testing set’ represents data ‘held back’ to test the method independently.
the Peccerillo & Taylor (1976) K2O^SiO2 diagram
that is used to classify volcanic arc lavas.
(2) To achieve this, an immobile element proxy for K2O
must replicate its incompatibility, its enrichment
above subduction zones and its enrichment during
assimilation and fractional crystallization. Th is the
most effective of the elements considered, being
2354
immobile up to lower amphibolite-facies metamorphism. Apart from arc volcanoes with a high shallow
subduction component, K^Th diagrams show that
Th replicates K2O well.
(3) The immobile element proxy for SiO2 must replicate
its steady change from basic to acid compositions and
its relative lack of variation in primary magmas.
HASTIE et al.
A Th-Co DISCRIMINATION DIAGRAM
100
100
H-K and SHO
H-K and SHO
10
1
CA
0.1
IAT
Th ppm
Th ppm
10
1
CA
0.1
IAT
B
BA/A
D/R*
B
0.01
BA/A
D/R*
0.01
70
60
50
40
30
20
10
0
70
60
50
Co ppm
40
30
20
10
0
Co ppm
Fig. 7. The Th^Co discrimination diagram. B, basalt; BA/A, basaltic
andesite and andesite; D/R, dacite and rhyolite ( indicates that
latites and trachytes also fall in the D/R fields). The field boundaries
are: B^BA/A ¼ (384, 001) to (24, 100); BA/A^D/R ¼ (23, 001) to
(7, 100); IAT^CA ¼ (70, 0245) to (0, 135); CA^H-K/SHO ¼ (70, 22)
to (0, 9).
No incompatible element or element ratio accomplished this. However, the slightly compatible element
Co did provide a good proxy, despite decreasing
rather than increasing during fractional crystallization. It has some advantages over Zr/TiO2 as a
proxy, being less affected by Fe^Ti oxide accumulation and zircon crystallization.
(4) The resulting Th^Co classification diagram acts as
the proxy for the K2O^SiO2 diagram. Fields drawn
on the basis of 85% probability contours demonstrate
that it is not possible to separate high-K calc-alkaline
from shoshonitic samples, nor basaltic andesite from
andesite or dacite from rhyolite samples with a high
degree of confidence. However, a ‘testing set’ of
samples not used in the original classification demonstrates that it can achieve c. 80% success rate in
separating tholeiitic from calc-alkaline from high-K
calc-alkaline plus shoshonite series, and in separating
basalt from basaltic andesite plus andesite from dacite
plus rhyolite rock types. Usefully, it is also topologically similar to the K2O^SiO2 diagram.
(5) This diagram is especially valuable for classification
of lavas within tropical regions because of the more
intense weathering, but works equally well with rocks
that have undergone hydrothermal alteration and/or
metamorphism. Retention of Co in iron and manganese oxyhydroxides and mafic metamorphic phases
such as amphibole and chlorite ensures that Co is
usually immobile. However, it should be stressed
that, as with other immobile trace element discrimination and classification diagrams, tests for element
immobility should always be carried out, and
Fig. 8. Th^Co classification of the Devils Racecourse lavas. Key as
in Fig. 2.
petrography and field evidence should be taken into
account in any application. Moreover, as with other
diagrams, the classification boundaries are not absolute but represent fields within which a particular
rock type or volcanic series is the most probable interpretation. Nevertheless, the Th^Co plot provides
a potentially useful projection for classifying altered
volcanic arc lavas.
AC K N O W L E D G E M E N T S
A.H. gratefully acknowledges receipt of a NERC PhD
Studentship (NER/S/A/2003/11215). We thank Iain
McDonald and Eveline DeVos for their analytical expertise in ICP-OES and ICP-MS at Cardiff University, and
MartinWolstencroft for his expertise in using the computer
program Generic Mapping Tools (GMT). A special thank
you must go to Arnott Jones, Ryan Ramsook, Ruth Liley
and Vencott Adams for providing valuable help with fieldwork and logistics in Jamaica, and to Mike Widdowson for
pointing me in the right direction at the beginning of the
project. Constructive reviews by Edward Lidiak, Kaj
Hoernle and Robert Trumbull significantly improved the
manuscript.
S U P P L E M E N TA RY DATA
Supplementary data for this paper are available at Journal
of Petrology online.
R EF ER ENC ES
Arculus, R. J. (2003). Use and abuse of the terms calcalkaline and
calcalkalic. Journal of Petrology 44, 929^935.
Becker, H., Jochum, K. P. & Carlson, R. W. (2000). Trace element fractionation during dehydration of eclogites from high-pressure
2355
JOURNAL OF PETROLOGY
VOLUME 48
terranes and the implications for element fluxes in subduction
zones. Chemical Geology 163, 65^99.
Brenan, J. M., Shaw, H. F. & Ryerson, F. J. (1995). Experimental
evidence for the origin of lead enrichment in convergent-margin
magmas. Nature 378, 54^56.
Burke, K., Coates, A. G. & Robinson, E. (1969). Geology of the
Benbow Inlier and surrounding areas, Jamaica. In: Saunders, J. B.
(ed.) Transactions of the Fourth Caribbean Geological Conference, Trinidad,
28 March^12 April, 1965, pp. 229^307.
Cann, J. R. (1970). Rb, Sr, Y, Zr and Nb in some ocean floor basaltic
rocks. Earth and Planetary Science Letters 10, 7^11.
Class, C., Miller, D. M., Goldstein, S. & Langmuir, C. (2000).
Distinguishing melt and fluid subduction components in Umnak
Volcanics, Aleutian Arc. Geochemistry, Geophysics, Geosystems 1, paper
number 1999GC000010.
Dosseto, A., Bourdon, B., Gaillardet, J., Alle'gre, C. J. & Filizola, N.
(2006). Time scale and conditions of weathering under tropical
climate: Study of the Amazon basin with U-series. Geochimica
et Cosmochimica Acta 70, 71^89.
Elliott, T. (2003). Tracers of the Slab. Inside the Subduction Factory. American
Geophysical Union, Geophysical Monograph 138, 23^45.
Floyd, P. A. & Winchester, J. A. (1975). Magma type and tectonic
setting discrimination using immobile elements. Earth and Planetary
Science Letters 27, 211^218.
Floyd, P. A. & Winchester, J. A. (1978). Identification and discrimination of altered and metamorphosed volcanic rocks using immobile
elements. Chemical Geology 21, 291^306.
Govindaraju, K. (1994). Compilation of working values and
samples description for 383 geostandards. Geostandards Newsletter
18, 1^158.
Hildreth, W., Fierstein, J., Siems, D. F., Budahn, J. R. & Ruiz, J.
(2004). Rear-arc vs. arc-front volcanoes in the Katmai reach
of the Alaska Peninsula: a critical appraisal of across-arc
compositional variation. Contributions to Mineralogy and Petrology 147,
243^275.
Hill, I. G., Worden, R. H. & Meighan, I. G. (2000). Yttrium: The
immobility^mobility transition during basaltic weathering. Geology
28, 923^926.
Hynes, A. (1980). Carbonitization and mobility of Ti, Y and Zr in
Ascot formation metabasalts, S.E. Quebec. Contributions to
Mineralogy and Petrology 75, 79^87.
Jackson, T. A. (1987). The petrology of Jamaican Cretaceous and
Tertiary volcanic rocks and their tectonic significance. In:
Ahmad, R. (ed.) Proceedings of a Workshop on the Status of Jamaican
Geology, Kingston 14^16 March 1984, Geological Society of
Jamaica, special issue. pp. 107^119.
Johnson, M. C. & Plank, T. (1999). Dehydration and melting experiments constrain the fate of subducted sediments. Geochemistry,
Geophysics, Geosystems 1, paper number 1999GC000014.
Keppler, H. (1996). Constraints from partitioning experiments on the
composition of subduction-zone fluids. Nature 380, 237^240.
Kessel, R., Schmidt, M. W., Ulmer, P. & Pettke, T. (2005). Trace
element signature of subduction zone fluids, melts and supercritical
liquids at 120^180 km depth. Nature 437, 724^727.
Kuno, H. (1968). Differentiation of basalt magmas. In: Hess, H. H. &
Poldervaart, A. A. (eds) Basalts: The Poldervaart Treatise on Rocks of
Basaltic Composition, 2. New York: Interscience, pp. 623^688.
Kurtz, A. C., Derry, L. A., Chadwick, O. A. & Alfano, M. J.
(2000). Refractory element mobility in volcanic soils. Geology 28,
683^686.
Le Bas, M. J., Le Maitre, R. W., Steckeisen, A. & Zanettin, B. (1986).
A chemical classification of volcanic rocks based on the total
alkali^silica diagram. Journal of Petrology 27, 745^750.
NUMBER 12
DECEMBER 2007
Le Bas, M. J., Le Maitre, R. W. & Woolley, A. R. (1992). The construction of the total alkali^silica chemical classification of volcanic
rocks. Mineralogy and Petrology 46, 1^22.
McCulloch, M. T. & Gamble, J. A. (1991). Geochemical and geodynamical constraints on subduction zone magmatism. Earth and
Planetary Science Letters 102, 358^374.
McDonald, I. & Viljoen, K. S. (2006). Platinum-group element geochemistry of mantle eclogites: a reconnaissance study of xenoliths
from the Orapa kimberlite, Botswana. Applied Earth Science
(Transactions of the Institution of Mining and Metallurgy B) 115, 81^93.
Miller, D. M., Goldstein, S. & Langmuir, C. (1994). Cerium/lead and
lead isotope ratios in arc magmas and the enrichment of lead in the
continents. Nature 368, 514^519.
Miyashiro, A. (1974). Volcanic rock series in island arcs and active continental margins. AmericanJournal of Science 274, 321^355.
Nakano, S. (1993). Ueno basaltic rocks I: Heterogeneous magmas
at two monogenetic volcanoes. Journal of the Japanese Association of
Mineralogists, Petrologists and Economic Geologists (Ganseki Kobutsu
Kosho Gakkai-Shi) 88, 272^288.
Nakano, S. (1994). Ueno basaltic rocks II: Chemical variation in the
Kiso Province, to the south of the Ontake volcano. Journal of the
Japanese Association of Mineralogists, Petrologists and Economic Geologists
(Ganseki Kobutsu Kosho Gakkai-Shi) 89, 115^130.
Pearce, J. A. & Cann, J. R. (1973). Tectonic setting of basic volcanic
rocks determined using trace element analyses. Earth and Planetary
Science Letters 19, 290^300.
Pearce, J. A. (1976). Statistical analysis of major element patterns in
basalts. Journal of Petrology 17, 15^43.
Pearce, J. A. (1982). Trace element characteristics of lavas from
destructive plate boundaries. In: Thorpe, R. S. (ed.) Andesites.
Chichester: John Wiley, pp. 525^547.
Pearce, J. A. & Peate, D. W. (1995). Tectonic implications of the composition of volcanic arc magmas. Annual Review of Earth and
Planetary Sciences 23, 251^285.
Pearce, J. A. (1996). A users guide to basalt discrimination diagrams.
In: Wyman, D. A. (eds) Trace Element Geochemistry of Volcanic Rocks:
Applications for Massive Sulphide Exploration. Geological Association of
Canada, Short Course Notes 12, 79^113.
Pearce, J. A. & Parkinson, I. J. (1993). Trace element models
for mantle melting: application to volcanic arc petrogenesis.
In: Prichard, H. M., Alabaster, T., Harris, N. B. W. &
Neary, C. R. (eds.) Magmatic Processes and Plate Tectonics (I. G. Gass
Memorial Volume). Geological Society, London, Special Publications 76,
373^403.
Peccerillo, R. & Taylor, S. R. (1976). Geochemistry of
Eocene calc-alkaline volcanic rocks from the Kastamonu
area, northern Turkey. Contributions to Mineralogy and Petrology 58,
63^81.
Regelous, M., Collerson, K. D., Ewart, A. & Wendt, J. I. (1997). Trace
element transport rates in subduction zones: evidence from Th, Sr
and Pb isotope data for Tonga^Kermadec arc lavas. Earth and
Planetary Science Letters 150, 291^302.
Rickwood, P. C. (1989). Boundary lines within petrologic
diagrams which use oxides of major and minor elements. Lithos 22,
247^263.
Robinson, E., Lewis, J. F. & Cant, R.V. (1972). Field guide to aspects of
the geology of Jamaica. International Field Institute Guidebook to the
Caribbean Island Arc System 1970. Washington D.C.: American
Geological Institute. 1^45, 1^4.
Roobol, M. J. (1972). The volcanic geology of Jamaica. In: Petzall, C.
(ed.) Transactions of the 6th Caribbean Geological Conference, Margarita,
Venezuela, 6^14 July 1971. pp. 100^107.
2356
HASTIE et al.
A Th-Co DISCRIMINATION DIAGRAM
Savov, I. P., Ryan, J. G., D’Antonio, M., Kelley, K. & Mattie, P. (2005).
Geochemistry of serpentinized peridotites from the Mariana
Forearc^Conical Seamount, ODP Leg 125: Implications for the
elemental recycling at subduction zones. |Geochemistry, Geophysics,
Geosystems| 6, paper number 2004GC000777.
Shervais, J. W. (1982). Ti^V plots and the petrogrenesis of modern and
ophiolitic lavas. Earth and Planetary Science Letters 59, 101^118.
Summerfield, M. A. (1997). Global Geomorphology. Singapore: Longman,
pp. 129^144.
Sun, S.-s. & McDonough, W. F. (1989). Chemical and isotope systematics of oceanic basalts: implications for mantle composition and
processes. In: Saunders, A. D. & Norry, M. J. (eds) Magmatism in
the Ocean Basins. Geological Society, London, Special Publications 42,
313^34542.
Trescases, J. J. (1973). Weathering and geochemical behaviour of the
elements of ultramafic rocks in New Caledonia. Bulletin of Bureau
of Mineral Resources, Geology and Geophysics, Canberra 141, 149^161.
Winchester, J. A. & Floyd, P. A. (1976). Geochemical magma type discrimination: application to altered and metamorphosed basic
igneous rocks. Earth and Planetary Science Letters 28, 459^469.
Winchester, J. A. & Floyd, P. A. (1977). Geochemical discrimination of different magma series and their differentiation products using immobile elements. Chemical Geology 20,
325^343.
Wood, D. A., Joron, J.-L. & Treuil, M. (1979). A re-appraisal of the use
of trace elements to classify and discriminate between magma
series erupted in different tectonic settings. Earth and Planetary
Science Letters 45, 326^336.
2357